Key Highlights

  • US data centers consumed 183 to 200 TWh of electricity in 2024, over 4% of total US electricity demand.
  • By 2030, US data center consumption could reach a midpoint estimate of 650 TWh, implying growth of more than 130%.
  • The International Energy Agency projects global data center demand near 945 TWh by 2030, expanding around 15% annually.
  • The United States and China are expected to account for nearly 80% of incremental global data center electricity growth.
  • More than half of projected US electricity demand growth through 2030 may be attributable to data centers, driven largely by AI inference.

A Structural Break in US Electricity Demand

For nearly 15 years, US electricity consumption was broadly flat. Efficiency gains, industrial restructuring, and moderate economic expansion kept total load growth subdued between 2005 and 2020.

That regime is shifting.

The U.S. Energy Information Administration projects record electricity demand in both 2025 and 2026, with total consumption rising from roughly 4,110 billion kWh in 2024 to more than 4,260 billion kWh by 2026.

This acceleration reflects a structural change in economic activity. Artificial intelligence infrastructure is emerging as a new, persistent source of industrial-scale power demand.

AI Inference as the Dominant Energy Driver

Initial market focus centered on AI model training, which requires large but episodic bursts of compute power.

The more durable shift lies in inference.

Inference refers to the continuous application of trained AI models across search engines, enterprise software, cloud platforms, and consumer devices. Unlike training, inference scales directly with user adoption and operates continuously.

Industry estimates suggest that over 80% of AI compute workloads are now devoted to inference. By the end of the decade, roughly three-quarters of AI-related electricity demand may stem from inference rather than training.

This distinction is central. AI is transitioning from a development phase to an embedded utility layer of the economy. Electricity demand rises accordingly.

The Scale of Data Center Expansion

US data centers consumed approximately 183 to 200 TWh of electricity in 2024, representing more than 4% of national electricity consumption. That level of demand is comparable to the annual electricity usage of mid-sized industrial economies.

By 2030, projections diverge widely. Conservative scenarios place US data center demand near 300 TWh. More aggressive estimates exceed 1,000 TWh. A midpoint projection of roughly 650 TWh implies growth of more than 130% from current levels.

Globally, the International Energy Agency expects data center electricity consumption to approach 945 TWh by 2030, expanding at around 15% per year. That growth rate is more than four times faster than most other end-use sectors.

The United States and China together are projected to account for nearly 80% of incremental global data center electricity demand.

In practical terms, data centers alone could represent more than half of total US electricity demand growth through 2030.

Electricity is becoming a binding constraint within the AI value chain.

Infrastructure Bottlenecks and Capital Allocation

Meeting this demand is neither immediate nor costless.

Transmission infrastructure in the United States faces interconnection backlogs and permitting delays. New high-voltage lines can require several years to approve and construct. Generation projects face similar regulatory and supply chain constraints.

Large-scale grid modernization is capital intensive. Estimates suggest hundreds of billions of dollars may be required this decade to expand transmission, upgrade substations, and enhance grid resilience.

In the near term, natural gas remains the primary source of dispatchable generation for new data center clusters. However, nuclear power is increasingly viewed as attractive baseload capacity due to its reliability profile.

The grid, rather than compute capacity, may emerge as the limiting factor for AI expansion.

Nuclear Generation as Baseload for AI

 

AI workloads require stable, 24-hour electricity supply.

Constellation Energy (NASDAQ:CEG) operates one of the largest nuclear fleets in the United States and has signed multi-year power purchase agreements with hyperscale technology firms seeking firm clean power.

Vistra Corp (NYSE: VST) combines nuclear and natural gas generation in high-demand markets, including Texas and PJM, positioning it within regions experiencing strong data center load growth.

Talen Energy (NASDAQ: TLN) maintains exposure to both nuclear and gas assets and has reported material revenue expansion tied to data center-related contracts.

For nuclear operators, AI demand introduces long-duration counterparties with structurally high electricity requirements.

Transmission and Grid Modernization

Generation must be matched with transmission capacity.

GE Vernova (NASDAQ: GEV) supplies generation and grid equipment across gas, wind, and nuclear markets, providing exposure to incremental capacity additions.

Quanta Services (NYSE: PWR) focuses on high-voltage transmission and grid construction, supported by a large and expanding project backlog.

MasTec (NYSE: MTZ) operates across energy and utility infrastructure, benefiting from clean energy and grid reinforcement initiatives.

In an environment of accelerating demand, transmission capacity may become strategically as important as generation assets themselves.

Utilities and Regional Load Growth

Regulated utilities in high-density data center markets are central to the transition.

NextEra Energy (NYSE: NEE) combines regulated utility operations with renewable and nuclear generation expansion. Its long-term infrastructure investment program reflects anticipated load growth.

Dominion Energy (NYSE: D) operates in Virginia, a leading US data center hub, and has reported substantial power demand discussions linked to hyperscale expansion.

Utilities must balance capital expenditure, regulatory oversight, and rate stability while accommodating rapid structural load growth.

Power Management and Grid Stability

Rising rack densities increase the importance of power quality and thermal management.

Vertiv Holdings (NYSE: VRT) provides critical power systems and cooling infrastructure tailored to AI workloads.

Eaton Corporation (NYSE: ETN) supplies electrical distribution systems integral to data center expansion.

Fluence Energy (NASDAQ: FLNC) focuses on grid-scale battery storage, supporting stability as AI-related loads introduce volatility.

Electricity as the Core Constraint in the AI Economy

The AI narrative has centered on semiconductors and software platforms. Yet the foundational input is electricity.

Data centers already account for more than 4% of US electricity demand. By the end of the decade, they could account for a majority of incremental load growth.

If inference adoption continues at scale, the AI cycle will increasingly resemble an infrastructure supercycle. Capital allocation, regulatory agility, and grid reliability will determine how smoothly the transition unfolds.

Electricity is no longer a background utility. It is becoming a strategic asset in the AI-driven economy.